Introduction
The partial substitution of fluorine anions for oxygen in molten silicates and aluminosilicates shows drastic effects on physical properties, such as fluidity and the melting point of systems1. Therefore, fluxes and slags containing fluoride were widely used in the hot metal treatment of the steel making process. Steel industries, however, are now forced to decrease the usage of fluoride to meet the recent environmental requirement. To develop a new process without fluoride usage, we must know the role of the fluoride in molten silicates and aluminosilicates quantitatively to design new fluxes and slags. Although many investigations into the physical and chemical properties of fluoride-contained melts have been conducted2-4, the roles of F on the properties of molten silicates and aluminosilicates are still obscure.
The Na2O-NaF-SiO2system was chosen as a first step of the investigation of the role of F in silicate melts, since it had a much wider liquid phase area compared with that of the CaO-CaF2-SiO2 system at the same temperature.
Consequently, it made it possible to vary the composition for the measurements so that the role of F could be more clearly identified. Generally speaking, industrial mould fluxes contained not only the alkaline-earth metal cation of Ca2+, but also the alkali metal cations such as Na+. So far, most of investigations of the role of F were mainly focused on the CaO-CaF2-SiO2 system. But a few works were conducted for the Na2O-NaF-SiO2system. Therefore, it can be useful to investigate the behaviour of F in the Na2O- NaF-SiO2 system for understanding the structure of the practical mould fluxes.
Raman spectroscopy is well suited to evaluating these long-range order structures of silicate melts. Therefore, in this investigation, Raman spectroscopic measurement was applied to investigate the effect of F. It is also well known that molecular dynamics calculation gives much information on the structural details such as bond angle
distributions or pair functions that are difficult to obtain by Raman measurements2. Therefore, in this investigation, molecular dynamics simulations were also employed to investigate the structural role of F- ions in molten alkali- silicate melts and simulated results were compared and discussed with the measured Raman spectra.
Raman spectroscopy measurements
The detail of Raman spectroscopic measurements was described previously3. Appropriate portions of reagent grade Na2CO3, SiO2and NaF powders were mixed together in an agate mortar in air and placed in a Pt crucible. The mixture was heated at 1073 K for 10.8 ks to decarbonate and then fired in a muffle furnace at 1673 K for 7.2 ks. The melted mixture was quenched into a water-cooled copper flat mould and then ground to form a powder. The prepared powder was melted again as a bead (only about 5 mg) at 1673 K on a Pt-Pt alloy thermocouple wire loop in air and quenched. Quenching was achieved by switching off the power to the hot thermocouple. The prepared sample was broken into several pieces of about 3x3x3 mm and used for measurement.
The initial compositions of the samples are presented in Table I. In the Table, Na2F2is used instead of NaF, since it is convenient to consider the substitution of 2F- to O2-(2F-
→ O2-). After the samples were prepared, their contents were analysed by the Energy Dispersion X-ray method (EDX). Their compositions were found to be almost the same as the nominal compositions within the resolution of EDX (about 1%).
Raman spectra of samples were obtained with an automatic micro-Raman system (JASCO NRS-2000).
Silicate samples were excited by the 514.5 nm line of a coherent Ar+laser (NEC GLG-2162) operating at 300 mW.
The CCD detector was maintained at a constant temperature of liquid nitrogen and all the Raman spectra measurements of samples were made at room temperature. Unpolarized
SASAKI, Y., URATA, H., and ISHII, K. Structural analysis of molten Na2O-NaF-SiO2system by Raman spectroscopy and molecular dynamics simulation.
VII International Conference on Molten Slags Fluxes and Salts,The South African Institute of Mining and Metallurgy, 2004.
Structural analysis of molten Na 2 O-NaF-SiO 2 system by Raman spectroscopy and molecular dynamics simulation
Y. SASAKI, H. URATA, and K. ISHII
Division of Materials Science and Engineering, Graduate School of Engineering, Hokkaido University, Japan
To determine the effect of F ions in the molten alkali silicate systems on their structures, quenched Na2O-SiO2-NaF systems were investigated by Raman spectroscopy and molecular dynamics simulation. The systematic increase of 1100 cm-1band intensity in the Raman spectra of the silicate melts accompanying the replacement of F for O provides the evidence for concomitant polymerization of melts. From the molecular dynamics simulation, it was confirmed that most of substituted F was mainly coordinated to Na+ions but not Si4+ions. These results were consistent with the mechanism in which F associated with otherwise network-modifying Na rather than with network-forming Si. Since F was associated to Na+ions, the replace of O ion by two F ions promote the polymerization of silicate melts.
Keywords: fluorine; molten flux; slag structure; molecular dynamics simulation; Raman spectroscopy.
spectra were measured for all samples and several spectra from each sample were measured for each composition. All spots used for micro-Raman analysis were examined for laser damage immediately following the runs and detectable damages were not observed.
Molecular dynamic simulation method
The molecular dynamics simulations were applied to the same compositions as measured by Raman spectroscopy and high F content compositions. The MD calculations were carried out by using the MXDORTHO program4. The pair potential used in these molecular dynamics simulations was the Busing approximation of Born-Mayer-Huggins form5of
[1]
where Rij is the distance between ions i and j, Zi is the charge of ion i, f0= 6.9472 x10-11N. aiand biare potential parameters that are related to the size and softness of the ith atom, respectively. The potential parameters of each atom were empirically determined to reproduce crystal structures and their melting points. The reliability of the molecular dynamics simulation results largely depended on the parameters employed. The reliability of parameter sets for the Born-Mayer-Huggins potential was well discussed by Sangster and Dixon18. The values of the parameter set that was used in this calculation are shown in Table II6. Hirao and Kawamura17established these values that satisfied the reproduction of various stable crystal structures of 30 compounds. These parameters so far have been successfully used by many workers to simulate molten oxide and fluorides.
For the calculation, three-dimensional periodic boundary conditions were applied to the basic cells containing about 3 000 atoms. The long-range Coulombic forces were calculated by the Edwald summation within the uncertainty of 0.01%. The equation of motion was integrated by the Verlet algorithm with the time step of 2.0 fs. All the calculation was carried out at constant pressure of 101 kPa.
The initial temperature was set to 4000 K to agitate the atoms and eliminate the effect of the initial distribution on the final structure. Namely, the distribution of the constituent atoms in the basic cells was set to be statistically random at the start of calculation. The temperature was then decreased to 3000, 2000, 1800 and finally 1673 K. The calculation process of changing temperature from one to another was carried out for 10 000 steps. The setting of temperature decreasing rate was 1 K
per 10 steps. It was confirmed that about 3 000 steps were enough to reach another equilibrium state by changing temperature. However, just to be safe, 10 000 steps calculation was adapted. All the calculated structural data were accumulated at each temperature. Based on these calculated data, cumulative coordination numbers, average distances between atoms, pair correlation functions and the anionic structural unit distributions (cf. ‘Linkage between si tetrahedrlas’ below) were calculated.
Results
Raman spectra
The change in Raman spectra of joins Na2O·SiO2- Na2F2·SiO2quenched melts with varying Na2O/Na2F2ratios are shown in Figure 1. There were an intense band of 950 cm-1 and a relatively weak band of 850 cm-1 in the Na2O·SiO2system without F. With an increase of F content, the intensity of about 1075 cm-1band gradually developed and also slightly shifted to higher frequency but that of 950 cm-1 was gradually decreased. The band 850 cm-1 also gradually decreased and finally disappeared. Assignments of main Raman bands in silicate melts have been basically established1,7,8: namely, the bands near 1100, 950 and 850 cm-1are due to symmetric -O-Si-O0 (e.g., sheet structural unit), -O-Si-O- (e.g., chain structural unit) and Si-O2- (monomer unit) stretch vibration respectively. The observed bands were essentially the same as that observed in binary silicate systems: namely, new bands did not appear due to U r Z Z e R f b b
a a R b b
ij i j ij o i j
i j ij i j
( )
= +(
+)
(
+ −) (
+)
( )
2/
exp /
Table I
Compositions (mol%) of samples studied in the system Na2O-Na2F2-SiO2
SiO2 Na2O Na2F2
Na2O·SiO2 50.00 50.00 0.00
10SiO2-9Na2O-Na2F2 50.00 45.00 5.00 10SiO2-8Na2O-2Na2F2 50.00 40.00 10.00 10SiO2-7Na2O-3Na2F2 50.00 35.00 15.00 10SiO2-6Na2O-4Na2F2 50.00 30.00 20.00 9(Na2O·SiO2)-0.5 Na2F2 48.65 48.65 2.70 8(Na2O·SiO2)-Na2F2 47.06 44.45 5.88
Figure 1. Raman spectra along the join Na2O·SiO2-Na2F2·SiO2
Z a (10 nm) b (10 nm)
Si 4 1.102 0.080
Na 1 1.260 0.080
F -1 1.565 0.085
O -2 1.626 0.085
Table II
Parameter set used in the present calculations
the F substitution. This means that F in these melts had a negligible effect on the vibration between Si-O. In other words, F ions might not be coordinated to the Si tetrahedral.
Thus, the spectra change due to the F substitution shown in Figure 1 simply means that the number of the sheet structural unit was increased, but that of the monomer unit was decreased. To examine the effect of F on the polymerization quantitatively, the intensity ratio of 1100 and 950 cm-1bands (I1100/I950) of the melts is plotted as a function of NaF content in Figure 2. The I1100/I950increased almost lineally with Na2F2content within the experimental scatters. Based on these results, it can be said that the polymerization was enhanced with an increase of F for the Na2O-NaF-SiO2 melts, at least within the present experimental compositions.
Molecular dynamics simulation
Information about O and F coordination for Si and Na ions in the simulated melts can be obtained from the pair correlation function (PCF) between ions iand j, gij(r), and the cumulative coordination number (CCN) curve N(r) given8by
[2]
[3]
where Nis the number of ions in the box, Vis the volume of the box, nijis the number of ions jabout ion iin the shell within distances r±∆r/2, < > is the time average, and δ is the particle density of an ion.
The individual calculated correlation functions for the Si- O and Si-F pairs and cumulative coordination numbers of O and F around Si at 1473 K in the simulated Na2O·SiO2- Na2F2·SiO2 (20 mol%) melt are shown in Figure 3 and Figure 4 respectively. The peak of PCF for Si-F pairs at around 0.172 nm was slightly asymmetric and had a small tail at longer distance. It suggests that the coordination of F around Si was not so tight or loosely bounded and differed from the case of Si-O bond. The Si-O and Si-F pair correlation functions yielded only the first peak at about 0.162 and that for the Si-F pair was found at 0.172 nm, and there was no other distinct peak at longer distance. This implies that, in the Na2O·SiO2-Na2F2·SiO2(20 mol%) melt,
only the first coordination sphere was distinguishable. In this first coordination sphere the O coordination number of Si ions was found to be 4 and that of F ions was about 0.1 from the cumulative coordination number curves. The coordination number of 4 for the Si-O pair simply indicated the Si tetrahedral formation. The coordination number of about 0.1 for the Si-F pair means that there exists one Si tetrahedral with an F ion coordinated per 10 Si tetrahedrals, if F plays a non-bridging role.
Linkage between Si tetrahedrals
Binary silicate melts are known to consist of mixtures of five distinct types of structural units that have NBO/Si =0, 1, 2, 3 and 4 (NBO/Si; nonbridging oxygens per silicon1).
In terms of a stoichiometric expression, these units were described as SiO44- (monomer), Si2O76-(dimmer), Si2O64-
(chain unit), Si2O52-(sheet unit) and SiO2 units. In the terminology commonly employed in NMR spectroscopy, these structures were referred to as Q0, Q1, Q2, Q3 and Q4 where the superscript referred to the number of bridging oxygen atom in a given structural unit. Thus SiO44- is a Q0, Si2O76-is a Q1, Si2O64-is a Q2, Si2O52-is a Q3, and SiO2is a Q4.
When oxygen atoms were found to exist within a separation of the first-nearest neighbour of a silicon, within 0.165 ±0.2 nm as shown in Figure 3, then it was assumed N r
( )
=4πδj∫
0rr g r dr2 ij( )
g r V
N N
n r r r
ij
i j
ij j
( )
=( )
∑
4π ∆2Figure 2. I1100/I950ratio as function of Na2F2
Figure 4. Pair correlation function for Si-F and cumulative F coordination numbers for Si
Figure 3. Pair correlation function for Si-O and cumulative O coordination numbers for Si
that Si-O bonds were formed, namely these oxygen atoms were determined to belong to the Si tetrahedral in the MD simulation. The interconnected relations between the Si tetrahedrals were evaluated based on the number of bridging oxygen atoms of each tetrahedral and consequently each Si tetrahedral was classified into five types of structural units (Q0to Q4). In this evaluation of the interconnected relation between Si tetrahedrals, the existence of F was neglected, since F was confirmed to play a nonbridging role.
Based on their classified types, their relative distribution ratios were calculated. The calculated distribution of Qi (i=1to4) as a function of NaF content are shown in Figure 5. With an increase of F content, the fractions of Q3 and Q4 increased, whereas Q2 and Q1decreased. Namely, the polymerization was promoted with increasing of F content just as observed in Raman spectra change. From the results of the Raman spectra measurements and MD simulations, within the present experimental condition, it can be said that the substitution of F for O in the sodium silicate melts enhanced the polymerization of silicate complex anions.
Discussion
From the Raman spectroscopic results and molecular dynamics simulation, it was confirmed that the substitution of F for O enhanced the degree of the polymerization of melts, at least with the composition of basicity of less than 1 and with NaF content of less than 30 mol%. Luth9also found the same tendency for the CaO-CaF2-SiO2 system and discussed the mechanism to enhance the polymerization by F substitution. His discussed mechanism is possibly applied to the Na2O-NaF-SiO2system. Namely, the mechanism consistent with polymerization accompanying the substitution of F for O was formation of loosely bounded Na-F complexes. Formation of such complexes would remove Na+from a network-modifying role so that the degree of polymerization was enhanced.
In other words, NaF in the sodium silicate melts worked just as a diluter so that the existence of NaF did not influence the relative distribution of the complex anions of silicate melts. Then, this mechanism predicts that melts along a join from NaF to a point on the Na2O-SiO2 join have the same distribution ratio of the silicate network units, except the relative amount of Na-F complexes.
Consequently, the Raman spectra of the silicate melts on
this line can be the same. To confirm this, Raman spectroscopic measurements were carried out for the Na2O·SiO2-NaF system as a function of NaF. The change in Raman spectra of join Na2O·SiO2-NaF with different NaF content is shown in Figure 6. Although the relative heights of bands were a little bit different from each other, it can be said that they were basically the same. The distribution of Qi(i=1 to 4) for the Na2O·SiO2-NaF system as a function of NaF was also calculated and shown in Figure 7. The relative amount of Qiwas reasonably independent from the increase of NaF content. These results support the previously mentioned polymerization mechanism.
Conclusions
Based on the investigation of the effect of F ions on the structure of molten Na2O-SiO2-NaF systems by Raman spectroscopy and molecular dynamics simulation, the following conclusions can be drawn:
Figure 5. Distribution of Qias a function of NaF
Figure 7. Distribution of Qias a function of NaF for the melts along the join Na2O·SiO2-NaF
Figure 6. Raman spectra along the join Na2O·SiO2-NaF
• The intensity of 1100 cm-1band in the Raman spectra, corresponding to highly polymerized silicate unit, was increased by the replacement of F for O
• Most of substituted F was found to be mainly coordinated to Na+ ions but not Si4+ions from the molecular dynamic simulation
• From the molecular dynamic simulation, a small amount of F was found to be coordinated to Si and the cumulative coordination number was found to be about 0.1
• From the molecular dynamics simulation and the change of Raman spectra, it was confirmed that the polymerization of melts was enhanced by the substitution of F
• The mechanism consistent with the polymerization development accompanying the substitution of F for O is formation of loosely bounded Na-F complexes.
Acknowledgements
This work has been partially supported by the Ministry of Education, Culture, Sports, Science and Technology under the Grants-in-Aid for Scientific Research, Grant No.13450303, and is gratefully acknowledged.
References
1. MYSEN, B.O. Structure and Properties of Silicate Melts, Elsevier, New York, 1988, pp. 247–253.
2. TSUNAWAKI, Y., IWAMOTO, N., HATTORI, T.,
and MITSUISHI, A.J. Analysis of CaO-SiO2 and CaO-SiO2-CaF2glasses by Raman spectroscopy. Non- Crystal. Solids, vol. 44 1981, pp. 369–378.
3. MOHRI, M., SASAKI, Y., and ISHII, K.
Comperative roles of Al3+and Fe2+ ions for network construction in sodium silicate melts. ISIJ Int., vol. 41 2001, pp. 410–415.
4. KAWAMURA, K. MXDORTH program. Japan Chemistry Program Exchange, No. 076.
5. BUSING, W.R. Interpretation of the crystal structure of Li2BeF4 in terms of the Born-Mayer-Huggind model. J. Chem. Phys., vol. 57 1972, pp. 3008–3010.
6. HIRAO, K. and KAWAMURA, K. Material Design using Personal Computer, Shokabo, Tokyo, 1994, pp. 52–54.
7. MCMILLAN, P. Structural studies of silicate glasses and melts-applications and limitations of Raman spectroscopy. Amer. Miner., vol. 69 1984, pp. 622–644.
8. KAWAMOTO, K., HORISAKA, T., HIRAO, K., and SOGA, N. A molecular dynamics study of barium meta-fluorozirconate glass. J. Chem. Phys., vol. 83 1985, pp. 2398–2404.
9. LUTH, R.W. Raman spectroscopic study of solubility mechanism of F in glasses in the system CO-CaF2- SiO2. Amer. Mineral., vol. 73 1988, pp. 297–305.
